Fetching from the wire…
01
02
03
04
05
06
07
08
Infra2026-08-11 · source-backed
TrendForce reporting via Tom's Hardware says prototype variants run 192GB or 256GB, down from the announced 288GB, with some configs using fewer than 16 stacks and substituting HBM4 for HBM4E. Driver is tightening HBM supply across SK hynix, Samsung, and Micron. At GTC 2025 configurations up to 1TB of HBM4E were on the table. Capacity planning consequence: a model that fits on four high-memory GPUs may need six or eight lower-memory ones, dragging along power, rack space, interconnect, and per-GPU licensing.
Each link below shares sources, entities, or timing with this story.
Internal survey data shows 81.5% of foundry staff and nearly 50% of the broader semiconductor division intending to leave within two years. The cause is bonus structure tied to division P&L: SK Hynix paid roughly $476,000 per employee, Samsung's memory division ~$400,000, and...
NVIDIA's Blackwell successor is in production ahead of schedule. The NVL72 rack (72 GPUs) delivers 3.6 exaFLOPS for inference, with 288GB HBM4 per GPU. NVIDIA claims 10x lower cost-per-token versus Blackwell. The Rubin CPX variant — purpose-built for million-token inference —...
The September 15 feature catalogs GPUs idling 50-80% of the time waiting on memory, then maps the contenders: Nvidia's $20B Groq acquisition producing an LPU with 500MB on-chip SRAM and 7x GPU memory bandwidth, a Cerebras WSE-3 deployment pushing GPT-5.3-Codex-Spark past 1,000...
Dario Amodei published "We Must Pace the Frontier" on September 12. Altman and Musk agreed within hours. Hassabis called the direction correct. By Monday morning the market had priced it. Nasdaq 100 futures fell 1.5% and S&P 500 futures 0.8%. Nvidia dropped 3%, AMD 5.7%, and A...
NVHBM, announced August 26, relocates NVIDIA's custom memory controller from the compute chip into the HBM base die, claiming up to 30% more bandwidth than standard HBM4E, 15% lower HBM power, and up to 25% more freed area on the XPU compute die (NVIDIA). Next-generation Train...
The August 23 Hot Chips session walked HBM1 through HBM4, with HBM3E's 128 banks per die doubling to 256 in HBM4, and showed a typical GPU package exceeding 12,000 square millimeters once eight HBM4 stacks are included. Roughly 3x as much silicon is consumed to deliver the sam...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.